Diagnosis of Psoriasis and Eczema Using Deep Features Based on Transfer Learning with Different Domains

Abstract

Misdiagnosis of skin diseases is a common occurrence worldwide. Psoriasis is a skin disease that has many similarities with other diseases, and its wrong diagnosis causes many problems in the treatment process. Misdiagnosis of this disease increases the length of treatment. The small number of medical images and dermatological databases makes examination and diagnosis difficult, so diagnosis using different images is very useful. Diagnostic methods using deep features based on transfer learning have received much attention in medical imaging today. This method reduces the dependency on matched data. Deep learning methods have been able to show their ability to recognize images well. Therefore, in this article, two data sets with different domains have been investigated using deep features based on transfer learning. In this research, the data of the first database was used for the training data and the data of the second database was used for the test, and this work was also set to the test mode for other data. In this article, by using convolutional neural network without manual intervention, deep features have been extracted and then three groups of diseases have been diagnosed. Accuracy results were calculated by 10-fold cross-validation method and reached 99.68% accuracy. This study shows that the proposed method differentiates skin diseases with acceptable accuracy.